earlyink
113 updated first flagged14 Oct 2026

The patch bill arrives, and BoA's 1,000 contractors set the price

Remediation labour and auto-fix tooling for flaws found by AI scanners. Banks and large enterprises pay.

Crowded · BoA hires 1,000 patch contractors

Written for: Investor or analyst

Next step

Thesis: AI-found vulnerabilities turn security budgets into remediation labour, and the winners will fix at contractor scale without the contractors. Confirm: another bank or insurer discloses remediation headcount or spend tied to AI scanners, or IT services firms cite it on earnings calls. Already building: OpenAI's Codex review on GitHub, with GitLab in preview, while the labour side is still contractors.

Why now

Our earlier card argued that security spending would shift from finding bugs to fixing them. This is the first public cost figure, and it comes from a systemically important bank. The same bank was just fined $1.2bn over AML failures, so regulators are watching and deferring fixes is not an option. Fixing is not automated yet. OpenAI's Codex code review is now generally available on GitHub, but the CTXGAP study shows coding agents leave failed attempts and risks out of their reports, which keeps humans in the loop.

What became possible

Frontier models now find vulnerabilities at scale. Anthropic's Mythos exposed enough flaws that Bank of America hired 1,000 contractors to patch them, and Argon found more than twice as many flaws as baseline models in Wiz testing.

Two ways in

  1. Fix-at-scale with regression proof

    Agents patch AI-found flaws and attach test evidence that behaviour is unchanged, so reviewers sign off in minutes.

    Customer
    Bank application-security teams
    Model
    Priced per remediated finding
  2. AI-native remediation services

    An outcome-priced services firm that replaces contractor headcount with agents plus a small review team.

    Customer
    Banks and insurers facing scanner backlogs
    Model
    Fixed fee per backlog tranche

Competition

Crowded

The exploitability-ranked queue has very little room left. Nucleus, Check Point, CrowdStrike, Synack and the newer remediation agents like Cogent and Reclaim all sell the "1,400 down to 50" story.

Full competition notes (13 competitors)

The exploitability-ranked queue has very little room left. Nucleus, Check Point, CrowdStrike, Synack and the newer remediation agents like Cogent and Reclaim all sell the "1,400 down to 50" story. CISA BOD 26-04 has also made exploitability-based ranking a compliance checkbox, which favors incumbents that already own the scanner and asset data. A new entrant would need a distinctly better exposure signal to compete there. The patch impact simulator is the more open idea. Today, testing before rollout means pilot rings (Ivanti, SecPod), checks the vendor runs on its own side (Adaptiva), or scripted app tests (Rimo3). Replaying captured production traffic against patched builds exists in research and in-house tools like PayU's ProdCast, but nobody I found sells it as a patch approval product for IT ops. The hard parts are real: capturing traffic for vendor-shipped line-of-business apps (Epic, core banking systems) rather than in-house code, stubbing calls that would hit outside systems, and handling PHI and PCI data in the replayed traffic. Those same hard parts are the moat. A focused entrant that proves "patch approved, no regression" evidence for a few regulated app stacks in hospitals and banks could win before Cogent-style agents or the patch vendors add it. Expect it to end up as an acquisition target for them rather than a standalone platform.

  • Cogent Security

    Raised a $42M Series A led by Bain Capital in February 2026. It builds AI agents aimed at the gap between finding vulnerabilities and fixing them, and the agents track remediation and confirm that patches are in place. This matches the core thesis almost exactly.

  • Reclaim Security

    Raised a $20M Series A from Acrew to automate cyber remediation, moving teams from identifying weaknesses to fixing them.

  • Tolmo (YC)

    YC company whose agents triage issues, verify impact and remediate on their own, escalating to humans only when needed.

  • Aikido Security (with Root.io)

    Launched agentic pentesting and fix validation in 2026, then bought Root.io, which supplies patched dependencies and container images without forcing major upgrades. The focus is on code and containers, not enterprise IT patching.

  • Nucleus Security

    Exposure management platform that ranks vulnerabilities by exploitability and business context rather than CVSS. Competes directly with the ranked patch queue.

  • Check Point Exposure Management

    Combines prioritization with exploitability validation, and markets itself on the claim that fewer than 1 in 12 critical vulnerabilities need urgent action. That is the same pitch as cutting 1,400 items down to 50.

  • CrowdStrike Falcon Exposure Management

    Large-vendor feature that uses adversary intelligence to rank exploitable exposures and drive fixes.

  • Synack (exposure validation)

    Pitches exploitability validation in response to CISA BOD 26-04, which ties remediation deadlines to evidence of exploitation instead of CVSS scores.

  • Rimo3

    Automatically tests whether applications still work after patches and updates are applied. This is the closest existing match to the patch impact simulator, though it runs scripted app tests rather than replayed production traffic.

  • SecPod Test and Deploy

    Adds a test stage before automated patch jobs. Rollout continues on its own once the customer's success criteria pass on test machines.

  • Adaptiva

    Autonomous patching. The vendor tests patches for install, compatibility and clean uninstall before publishing them to its library, but this happens on the vendor's side, not in the customer's environment.

  • Ivanti Endpoint Manager / DSM

    Patch rollout projects with pilot rings and dependency or incompatibility checks before wide deployment.

  • PayU ProdCast

    In-house tool that replays real production traffic against new code before deploy. It shows the traffic-replay method works, but it is not sold as a patch approval product.

Checked 2026-10-14

Risks

  • This may be a one-time backlog surge rather than recurring spend.
  • IT services incumbents may capture the budget by default.
  • Auto-fixes can introduce regressions that agent reports leave out.

Strong market pull and team fit; open space is the weak point (5).

tech 6 · pull 8 · space 5 · fit 8 · build 6

Exposed

Companies that could benefit or be threatened. News and context, not investment advice.

Momentum of its main topic

168 developments this week +53 vs last week
21 Sep: 0 developments22 Sep: 0 developments23 Sep: 0 developments24 Sep: 0 developments25 Sep: 0 developments26 Sep: 0 developments27 Sep: 0 developments28 Sep: 0 developments29 Sep: 0 developments30 Sep: 0 developments1 Oct: 0 developments2 Oct: 0 developments3 Oct: 0 developments4 Oct: 0 developments5 Oct: 0 developments6 Oct: 0 developments7 Oct: 0 developments8 Oct: 32 developments9 Oct: 26 developments10 Oct: 13 developments11 Oct: 11 developments12 Oct: 14 developments13 Oct: 19 developments14 Oct: 24 developments15 Oct: 35 developments16 Oct: 31 developments17 Oct: 15 developments18 Oct: 14 developments19 Oct: 23 developments20 Oct: 26 developments
Stage: Coverage

Sources

  1. Bank of America hired 1,000 contractors to patch software flaws exposed by Anthropic's MythosCNBC
  2. Using Gemini 4 Argon to protect against Cyber attackersGoogle DeepMind
  3. BoA fined $1.2bn over "egregious" AML deficienciesFinextra
  4. Code review is no longer the bottleneckOpenAI
  5. What Coding Agents Leave Unsaid: Measuring Task-State Gaps Between Execution and ReportsarXiv

Status history

  • 2026-10-14 Flagged as buildable